I know that a htm is not a brain in itself, but can be a recording of a single
object from the point of view of the brain, for variable recreation of an
object oriented record, to test mathematical randomization. (which calculus
itself suits the actual being, better.)
Imagine if you had a brain, and in it were 10 htms for the robot to see 10
different things. the input to the htms is the vector distances of their poses
of their moving parts, the main robot is a randomizing calculus pipe (similar
to karl sims evolved creatures) he uses these relating variables, for him to
test mutations in novel ground, the records inside the htms dont change but are
classified by the input variables (the distances and velocities of the moving
parts between the objects) themselves, to generate the distance and velocity of
the future. (the future behaviour of the animal)
Im going to implement soon, im stuffed with the fact the environment and amount
of objects in it must never change - but the dynamic polymorphic environment
is definitely on the books if my idea works.
_______________________________________________
nupic mailing list
[email protected]
http://lists.numenta.org/mailman/listinfo/nupic_lists.numenta.org
